Fig. 3: RR population provides accurate decoding of RR position.
From: A flexible hippocampal population code for experience relative to reward
a, Example decode of animal’s RR position from activity of RR cells. Decoder is trained on trials before the switch and tested on held-out trials before the switch (top) or tested on trials after the switch (bottom). b, Mean decode score compared to the mean of shuffles per session (n = 77 sessions, 11 mice, seven switch days), for each subpopulation of cells classified by remapping type. ***P = 7.54 × 10−14, two-sided Wilcoxon signed-rank tests, Bonferroni-corrected for multiple comparisons. Effect sizes: test before: RR, 2.6; TR, 3.3; non-RR, 3.2; test after: RR, 2.7; TR, −0.5; non-RR, −0.2. c, Top three rows: mean ± s.e.m. decode score from each subpopulation (across n = 11 mice) binned as a function of RR position, after z-scoring to each session’s shuffle, trained on trials before and tested on trials after the switch. Bottom row: distribution of reward delivery positions on the first (left) or last (right) switch day. Note that decoding accuracy from RR cells extends far beyond the reward location.
